A scalable, high-precision, and low-noise detector of shift-invariant image locations

A scalable, high-precision, and low-noise detector of shift invariant locations in grayscale images is presented. It leads to a wide range of novel image-to-‘data structures’ processing algorithms. Experiments with a single algorithm of this range prove that (i) the output structures convey great amount of semantically relevant information about the original image; (ii) this information can be successfully extracted and used in subsequent applications.

Research highlights
► A novel detector of shift-invariant image location is presented.
► Experiments show its high-precision, robustness, and low-noise.
► It leads to a wide range of novel image-to-‘data structures’ processing algorithms.
► An image constructed inversely by the structures resembles strongly the original.
► Analysis of the complexity and a comparison with Canny’s detector is presented.